A blank is ranked LAST — and rank numbers don't matter¶

The single most common Ranked Robin misconception, and the ballot-instruction ambiguity that feeds it. If you take one thing from this page: Ranked Robin reads the preference ORDER on each ballot, never the numeric rank label. A candidate you leave blank sits below everyone you ranked; whether you wrote your worst choice in the "last tier" or left them blank changes nothing.

The question (a real one)¶

It comes up every time a newcomer meets a Ranked Robin ballot with, say, five ranking bubbles:

"If I leave a candidate blank, are they ranked 5th, or 6th? The instructions say 'ranked last' — but which place is last?"

And close behind it, the deeper worry:

"If one voter ranks a candidate 5th and another leaves them blank (6th), doesn't that give the 5th-place candidate a win over the blank one — just from the rank numbers?"

Both questions dissolve once you see how the count actually works. (This is not hypothetical — it's the exact confusion, and its resolution, from an Equal Vote community thread. Worth teaching because three experienced advocates in that thread first assumed the ballot always has exactly five tiers. It doesn't.)

Answer, part 1 — a blank is ranked below everything you ranked¶

A Ranked Robin ballot is just an instrument for capturing one voter's order of preference. "Ranked last" doesn't point at a specific bubble (5th, 6th, 100th) — it means placed after every candidate you did rank. Leave Cara and Dan blank, rank Ada and Ben, and your ballot says:

Ada = Ben  >  (everyone I ranked)  >  Cara = Dan

So a blank is a real signal: it ranks that candidate below all your ranked choices. The number on the last bubble is irrelevant — 5 tiers or 6 or as many as there are candidates, "blank" always means under all of them.

Answer, part 2 — the count uses order, not rank numbers (this is not Borda)¶

Here's the misconception worth killing on sight. Ranked Robin is not Borda — it does not add up rank values. A candidate does not earn "points" for sitting in a higher-numbered tier, and rank numbers never transfer between ballots.

What Ranked Robin actually does: for every pair of candidates, count how many voters rank A above B and how many rank B above A. Most head-to-head wins takes the seat (that's Copeland; ties broken by margin, then lot). See Ranked Robin (the method).

So the "5th vs 6th" fear is a category error. Whether voter 1 writes their worst choice in the last tier or leaves them blank, that candidate lands last on voter 1's ballot either way — and contributes identically to every pairwise tally. The numeric gap between "5th" and "blank" is a distinction without a difference.

See it run¶

Four candidates — Ada, Ben, Cara, Dan — three voters. Dan is ranked explicitly last by voter 1, and left blank by voters 2 and 3. Watch Dan get treated identically (dead last) either way:

Ada > Ben > Cara > Dan (Dan written last)
Ada > Cara > Ben (Dan blank)
Ben > Ada > Cara (Dan blank)

Round-Robin — every pair, head-to-head (For – Against):
   Ada   beats Ben    2 – 1
   Ada   beats Cara   3 – 0
   Ada   beats Dan    3 – 0
   Ben   beats Cara   2 – 1
   Ben   beats Dan    3 – 0
   Cara  beats Dan    3 – 0

Win–loss record — Copeland score = wins + ½·ties:
    #  Candidate  W–L–T  Copeland  Margin  Beats
    1  Ada        3–0–0         3      +7  Ben, Cara, Dan
    2  Ben        2–1–0         2      +3  Cara, Dan
    3  Cara       1–2–0         1      -1  Dan
    4  Dan        0–3–0         0      -9  —

Winner — Ranked Robin (RCV-RR): Ada
   beats every opponent head-to-head — the Condorcet winner.

Look at Dan's column: "Ada beats Dan 3 – 0" — all three voters are counted as preferring Ada to Dan, including the two who left Dan blank. Blank = below every ranked candidate, exactly as if written last. Dan loses every pairwise 3–0 and finishes last; Ada beats everyone (the Condorcet winner) and wins outright. The rank numbers never entered the arithmetic.

Want the whole count? Full LH report → rr_blank_is_last_c4_b3_tabulated.txt · source → rr_blank_is_last_c4_b3.yaml. Independently cross-checked with pref_voting's Copeland (AGREE, unique winner).

The one subtlety that is real — multiple blanks are tied¶

Ranking your single worst choice "last" and leaving them "blank" are the same. But leaving two or more candidates blank is not the same as ranking them in some order — the blanks are all tied for last with each other. Two blanks means neither beats the other; you cast a pairwise tie between them (an Equal Support contest), not a preference. If you actually prefer one of your bottom two, you must rank them to say so.

That's the honest thing a ballot instruction has to convey: blanks rank below everyone you ranked, and blanks tie each other.

So what should the ballot say?¶

The trap wording is anything tier-specific — "ranked last (i.e. 6th)" bakes in an assumption that the ballot has exactly five tiers. It doesn't have to: five bubbles is a common, tested default, but a Ranked Robin contest can offer four, six, or as many tiers as there are candidates. The moment the tier count differs, "6th" is wrong.

The universally-correct instruction is order-based, matching how the count works and how this repo already frames the no-preference bucket as Equal Support:

Candidates left blank are ranked below all others (and tie one another).

No tier number, so it survives any ballot and any contest — the same discipline that keeps our Equal Support legend from hard-coding a candidate count. (Equal Vote's own preference is poll-testing the exact public wording before adopting it; the principle above is what the wording needs to preserve.)

See also: Ranked Robin (the method) · why it isn't Borda — the glossary · Strict vs. weak ranks (why equal ranks / ties are first-class here) · Summability. Up: Ranked Robin concept pages.

file: rr_blank_means_last.md¶